ACL and --without-all

ACL and --without-all

[Dmitry Antipov]

Could you please fix ACL detection to respect --without-all?

Also I believe that it should be --with/--without, not --enable/--disable.

Thanks,
Dmitry

Re: ACL and --without-all

[Paul Eggert]

Dmitry Antipov wrote:
> Could you please fix ACL detection to respect --without-all?

That'd be easy enough, something like the patch at the end of this email.

> Also I believe that it should be --with/--without, not --enable/--disable.

Other GNU packages use --enable/--disable, though;
it's better to be consistent.

Do you still think --without-all should imply --disable-acl?
If so, I can install this.

=== modified file 'configure.ac'
--- configure.ac	2013-08-15 20:37:35 +0000
+++ configure.ac	2013-08-16 05:47:53 +0000
@@ -4621,6 +4621,7 @@
 LIBS="$LIB_PTHREAD $pre_PKG_CONFIG_LIBS"
 gl_ASSERT_NO_GNULIB_POSIXCHECK
 gl_ASSERT_NO_GNULIB_TESTS
+test "$with_features" = no && : ${enable_acl=no}
 gl_INIT
 CFLAGS=$SAVE_CFLAGS
 LIBS=$SAVE_LIBS

Re: ACL and --without-all

[Stephen J. Turnbull]

Paul Eggert writes:
 > Dmitry Antipov wrote:

 > > Also I believe that it should be --with/--without, not --enable/--disable.
 > 
 > Other GNU packages use --enable/--disable, though;
 > it's better to be consistent.

The --enable/--with distinction is unnecessary and hard to understand.

XEmacs simply makes --with-foo and --enable-foo equivalent by
automatically defining both the --with-foo option and the --enable-foo
option, automatically setting the corresponding with_foo variable when
the --enable-foo option is used, and ignoring the enable_foo variable.
When it makes sense to have a --enable-foo option and a --with-foo-lib
option, that's exactly what we do.

Works great, for developers and users.  No backward-compatibility
complaints about this system in 10 years, either.  Apparently it never
broke ./configure --recheck or anybody's scripts.

Re: ACL and --without-all

[Dmitry Antipov]

On 08/16/2013 11:38 AM, Stephen J. Turnbull wrote:

> Paul Eggert writes:
>   > Dmitry Antipov wrote:
>
>   > > Also I believe that it should be --with/--without, not --enable/--disable.
>   >
>   > Other GNU packages use --enable/--disable, though;
>   > it's better to be consistent.
>
> The --enable/--with distinction is unnecessary and hard to understand.

I always treat this distinction as follows.

--with-X means that X is the feature 'of it's own', externally implemented
(mostly as a library), more or less independent from other features enabled
with other --with= options, and makes sense outside of the Emacs code.
That's why --with-dbus, --with-png and so it should be --with-acl (IMHO).

--enable-X means that X is the feature of our own (Emacs) code, and mostly
affects everything enabled with --with option. For example, --enable-debug
turns eassert into something useful everywhere, --enable-link-time-optimization
causes linking with lto1 backend for all object files, etc.

Dmitry